I wonder if the problem is in smoutput.  A couple of weeks ago I had a J 
crash when I put smoutput inside a recursive function and it generated a 
lot of typeout.

I don't know of any problems with -/"1 though.

Henry Rich

On 5/3/2011 7:28 PM, bill lam wrote:
> I can also reproduce the problem.  I found the box operation such as
> smoutput savedata=: cdir;bdysegs
> is sufficient to mutate the value of bdysegs
>
> bdysegs was also mutated after rank 1 reduction:
>   -/"1 bdysegs
>
> Втр, 03 Май 2011, Raul Miller писал(а):
>> On Mon, Apr 25, 2011 at 11:31 PM, Arthur Anger<[email protected]>  wrote:
>>> I think I have seen successive requests for the value of one of these
>>> variables
>>
>> I can reproduce this problem.
>>
>> It's obscured, somewhat by the use of ? in your definition of nreg.
>> But nreg will only ever be 1 or 2.  When it is 2, the code fails
>> before assigning bdirs.  When it is 1, the problem becomes visible.
>>
>> It looks to me as if the value behind bdysegs is sometimes being
>> mutilated by the parser.
>>
>> Specifically, I had defined:
>> cmoutput=: smoutput@,&<
>>
>> And then made some changes to your code, including:
>> smoutput savedata=: cdir;bdysegs
>> 'bdirs' cmoutput bdirs=: (2* * cdir),: - -/"1 bdysegs
>>
>> And I would get sessions that had stuff like this in it:
>>
>> +---+---+
>> |0 2|0 2|
>> |   |2 2|
>> +---+---+
>> +-----+-------------------------------+
>> |bdirs|         0                    2|
>> |     |_159704896 _9223372036854775807|
>> +-----+-------------------------------+
>> |index error: csnake
>> |   turnal=:0>.(bi    {bturns)*travel*1+sep21
>>     savedata
>> +---+--------------+
>> |0 2|159884768 4096|
>> |   |        1    1|
>> +---+--------------+
>>
>> But when I replaced cmoutput with smoutput@,&<  the value in savedata
>> matched the double box of numbers displayed just before the index
>> error.
>>
>> Thus, I am suspecting an uninitialized variable in the C
>> implementation of the parser.
>>
>> --
>> Raul
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to